Effects of Incentives vs Normative Feedback on Bio-waste Sorting in the Field

INTRODUCTION: The treatment and randomization unit in the study will be households located in the areas of the city of Palma (Spain) where individuals must use a personal card to use the bins meant for bio-waste sorting. Hence, each bin use is recorded and associated with one household, allowing us to track the participating households' use of the bio-waste sorting bins.

RECRUITMENT: Participants were recruited first between the 27th of February 2023 and the 30th of June; a second recruitment took place between 30th November 2023 and 25th January 2024. Recruitment was carried out face-to-face by directly approaching people in the street, by the internet through advertisements in social media, and by placing flyers in the city. The participants were informed that the study would be about bio-waste sorting behavior and that they would receive messages by WhatsApp or SMS; in the first round recruitment was facilitated by offering those who signed up to win one of four prizes of €250 each (only one individual per household could sign up). The certified random draw of the winners was performed by the independent firm "Easypromos" on the 27th of July. Each participant signed an informed consent form and answered a questionnaire that collected basic demographic data (number and ages of people in the participant's household, annual income, contact information, …) as well as data on social norms, personal norms, and relationship with nature (beliefs about average environmental sensitivity in the population, motivations to recycle, ...). Among those who signed up, only residents of Palma living in a neighborhood with the new electronic bio-waste sorting bins, in possession of the card required to open the bio-waste sorting bins, and who provided a valid cell phone number were included in the final sample.

2024-02-10

2024-03-11

Dummy variable that indicates for each week whether the household used the bio-waste sorting bin (1) or not (0)
Proportion of weeks that each household uses the bio-waste sorting bin at least once.

"Proportion of weeks that each household uses the bio-waste sorting bin at least once in each week" is computed by aggregating the "Dummy variable", indicating for each week whether the household used the bio-waste sorting bin (1) or not (0), over households and weeks.
This measure provides us with a measure of the average proportion of weeks that households recycled during the studied period which captures systematic recycling participation.

This RCT consists of a field experiment about bio-waste recycling in the city of Palma (Spain). Citizens of Palma must scan a card that identifies them in order to use the bins meant for bio-waste sorting. Such bins are available on streets throughout the city, next to the bins for recycling of glass, plastic, and paper. The city's waste management company will provide the research team with data on the bio-waste sorting behaviors of the 1,329 households who signed up to participate in the study between 27th February and 30th June 2023, and between 30th November 2023 and 25th January 2024. The goal is to study how the participants' use of the bio-waste sorting bins is affected by intrinsic (Kantian) and extrinsic (incentives) motives. The participants will be allocated to one of three treatments in two different waves.

Not available

Household level randomized by a computer. The assignment to the treatment arm is done through the "randtreat" command in Stata stratifying by recruitment source (Form, Twitter, street, 2nd recruitment wave) using setseed(51). The assignment to the "texting wave" is done through the "randtreat" command in Stata stratifying by treatment and recruitment source (Form, Twitter, street, 2nd recruitment wave) using setseed(2)

The randomization unit is the household.

No

1,329 households.

Weekly behavior of the 1,329 households.

443 households per treatment arm.

According to pre-intervention data, the average household in our study uses bio-waste sorting bins with a weekly frequency of 0.1876207, showing a standard deviation of 0.324002. For a power analysis of 80%, given the sample size we intend to have, we guarantee, at least, a minimum detectable effect of 0.0610 in the use frequency at 95% confidence and 0.0745 at 99% confidence.